Msg#: 4131295 posted 6:38 am on May 13, 2010 (gmt 0)
Thanks. Yes that is where we come in actually - and are tasked with manually giving a 1-5 rating for the (subjective) strength of the link. I just find it rather potentially accuracy blurring/denting that their is such a huge divide between the total number of webpages Google and Yahoo! claim to have crawled - a billion pages here or there would constitute a quantitatively if not qualitatively disparate data source...?
Msg#: 4131295 posted 8:09 am on May 13, 2010 (gmt 0)
Good luck with that 1 to 5 ranking project. You'll need a relatively complex formula to even come close to weighting backlinks the way Google seems to. I'm still boggled by it regularly. I can see some patterns that seem to hold true, but then I run into those mysterious exceptions.
One thing Google has told us is that on their back end, they can zero out the effect of links on a site-wide basis, a page-wide basis, or even at the level of an individual link. And if you're analyzing a huge pile of backlinks, it will pretty much be a wild guess as to whether one of those three possibilities is happening.
Even if you have your own independent link graph of a big chunk of the web, I still doubt you could reverse engineer everything that Google does to modify link strength. For example, there are historical factors - and they vary by the kind of site and keyword area.
Some links GROW in strength when they've been in place for a longer time. Some links WEAKEN in strength when they've been in place for a longer time. If a site has a high trust value with Google, then its links may carry more clout - especially if they use keyword anchor text.
And it seems to me that even pure PageRank calculations are now modified by where on the page a link appears. Check out this newly granted patent [seobythesea.com] which was filed back in 2004. The original idea of a random surfer is replaced by a "reasonable" surfer. In other words, a link that a page visitor is less likely to click on (for any number of reasons) will vote less power to its target website.
Msg#: 4131295 posted 8:16 am on May 13, 2010 (gmt 0)
So as their techniques and inclusion/exclusion criteria become more sophisticated and we move ever closer toward a form of legitimate artificial intelligence is their even scope for the endeavor of human implemented search engine optimization beyond perhaps only the most basic of guidelines regarding best practices in page, link and content lay-out?